--- license: other language: - en pipeline_tag: visual-question-answering tags: - medical-image-analysis - visual-question-answering - multimodal - reasoning - arxiv:2602.01081 datasets: - zhtstar/MedAD-38K ---
# MedAD-R1 ### Consistency-Reinforced Policy Optimization for Interpretable Medical Anomaly Detection

arXiv GitHub Hugging Face dataset ModelScope dataset Hugging Face model ModelScope model

> **Access.** Model access requests are reviewed while the paper is under review. We plan an open release after acceptance, subject to applicable permissions. ## Model overview **MedAD-R1** is a compact multimodal model for interpretable medical anomaly detection. It answers medical-image questions while producing structured reasoning that is trained to remain grounded in the image and consistent with the final answer. The model is trained on **MedAD-38K**, which contains **38,422 medical images** and **130,146 VQA instances**. Training combines supervised **Cognitive Injection** with **Consistency Group Relative Policy Optimization (Con-GRPO)**.
MedAD-R1 framework
## Reported performance | Model / training | MedAD-38K accuracy | Reasoning-answer consistency | External accuracy | | :-- | --: | --: | --: | | Qwen3.5-0.8B, zero-shot | 76.43 | 92.12 | 68.12 | | Qwen3.5-0.8B, SFT only | 90.89 | 94.35 | 88.36 | | **MedAD-R1, SFT + Con-GRPO** | **95.12** | **97.26** | **90.09** | See the paper for the full evaluation protocol and baseline comparison. This research checkpoint is intended for research use and has not been validated for clinical diagnosis or patient care. ## Project resources | Resource | Links | | :-- | :-- | | Paper | [MedAD-R1: Consistency-Reinforced Policy Optimization for Interpretable Medical Anomaly Detection](https://arxiv.org/abs/2602.01081) | | Code | [GitHub](https://github.com/zhtstar/MedAD-R1) | | Dataset: MedAD-38K | [Hugging Face](https://huggingface.co/datasets/zhtstar/MedAD-38K) · [ModelScope](https://modelscope.cn/datasets/zhtstar/MedAD-38K) | | Model: MedAD-R1 | [Hugging Face](https://huggingface.co/zhtstar/MedAD-R1) · [ModelScope](https://modelscope.cn/models/zhtstar/MedAD-R1) | ## Citation ```bibtex @article{zhang2026medadr1, title = {MedAD-R1: Consistency-Reinforced Policy Optimization for Interpretable Medical Anomaly Detection}, author = {Zhang, Haitao and Wang, Yingying and Wang, Jiaxiang and Xu, Haote and Zhang, Hongyang and Chen, Yirong and Huang, Yue and Ding, Xinghao}, journal = {arXiv preprint arXiv:2602.01081}, year = {2026} } ```